Soil moisture analysis switzerland

TOC

Comparison of modelled soil moisture (PREVAH, ERA5, ERA5-Land, Lisflood, UERRA) with in-situ soil moisture from the SwissSMEX (SWMX) dataset.

Datasets

Model data were interpolated to represent soil moisture at the same depth (50 cm) as the in situ measurement.

Models

PREVAH represents a high-resolution, locally optimised hydrological model and acts as a reference for this comparison.

Reference

SwissSMEX in-situ measurements: Mittelbach and Seneviratne (2012) DOI: 10.5194/hess-16-2169-2012

We selected 6 stations with long and uninterrupted time-series:

Map with SwissSMEX sites

Methods

Analysis

The following section shows several plots for each of the six selected SwissSMEX sites: time-series plots covering the full period of available data; time-series plots just for the years 2012 and 2018; Scatterplots which show a direct comparison between in-situ and modelled soil moisture.

The table summarizes the accuracy of modelled soil moisture, in terms of R and RMSE, with respect to the SwissSMEX measurements. The average in the last column refers to the average soil moisture computed over all stations.

R

-------- Payerne Plaffeien Wynau Chamau Reckenholz Taenikon Average
PREVAH 0.705 0.678 0.749 0.437 0.642 0.624 0.623
ERA5 0.617 0.592 0.680 0.442 0.614 0.574 0.647
ERA5-L 0.641 0.536 0.745 0.455 0.585 0.547 0.642
LISFLOOD 0.736 0.544 0.712 0.194 0.794 0.778 0.733
UERRA 0.721 0.678 0.634 0.541 0.539 0.657 0.577

RMSE

-------- Payerne Plaffeien Wynau Chamau Reckenholz Taenikon Average
PREVAH 0.803 0.842 0.760 1.164 0.908 0.906 0.914
ERA5 0.910 0.928 0.837 1.064 0.905 0.938 0.870
ERA5-L 0.865 1.005 0.749 1.059 0.934 0.964 0.878
LISFLOOD 0.752 0.859 0.775 1.274 0.670 0.662 0.741
UERRA 0.761 0.835 0.917 0.992 0.961 0.849 0.963

This first overview shows that correlations are relatively low in general. Interestingly, higher spatial resolution does not always lead to higher accuracies. Performances are relatively similar across the tested models.

Individual Stations

Payerne

Plaffeien

Wynau

Chamau

Reckenholz

Taenikon

Average Conditions

The plots below shows the time-series of average soil moisture

Time-series

Scatterplots

Drought Sensitivity

Next we want to evaluate the ability of different soil moisture models to detect drought. For this purpose we classify drought based on three SMA thresholds (<-1, <-1.5, <-2) and evaluate how well the models can capture drought events. We used the following metrics to quantify the classification accuracy:

Where P is the true number of "positive" classifications (i.e. drought) and N the number of "negative" classifications (i.e. no drought).

SMA < -1

SMA < -1.5

SMA < -2.0

Observations